Automatic Generation of Multiple-Choice Fill-in-the-Blank Question Using Document Embedding

摘要

Automatic question generation is a challenging task [11] that aims to generate questions from plain texts, and has been widely and actively researched in various fields. Generated questions can be used for educational purposes, largely for mid-terms, final exams, and also for pop quizzes. In this paper, we propose a novel similarity-based multiple choice question generation model without any pre-knowledge or additional dataset.